Study Of Composite Load Model Parameter Identification And Load Characteristics Classification And Synthesis

Power system digital simulation has become the primary tool for power system planning, operation, control and decision making. Load model is one of the crucial factors that influence the accuracy and reliability of the simulation results. Owing to random time variation, composition diversity, geographic dispersion and non-linear characteristics of the load, the load modeling is still a recognized problem of power system both at home and abroad.Load model structure has always been the most basic and critical problem of load modeling, which has attracted people's attention widely. Aggregate power load consists of various static load and dynamic load, therefore, it is hard to describe aggregate power load characteristics accurately by only using singular static load model or dynamic load model, while, adopting composite load model combing static load model and dynamic load model is becoming a trend. Noticing that Existing mechanism composite load model are too complicated and the parameters are not easy to be identified, this paper adopts improved ZIP/Exponential+Difference Equation Model to describe load characteristics, ZIP/Exponential and Difference Equation separately represent the static load and dynamic load of aggregate power load. Fault recording data obtained from dynamic simulation test is taken as load modeling data. According to model response feedback identification, the objection functions corresponding to parameter identification are established. PSO (Particular Swarm Optimization) algorithm is adopted to solve the optimization problem to identify the parameters of improved ZIP/Exponential+Difference Equation Model. Results show that this composite load model is very simple, its parameters are easy to be identified, it can describes the characteristics of aggregate power load accurately, so it can be a good choice for load model.Wide area power system has numerous load nodes, it is difficult to build load model for every node. A practical method is to classify load nodes into several kinds, pick out typical load node and build its load model, then generalize it to other load models of the same kind. A new method based on SOM (Self-Organizing Feature Map) neural network for Classification and synthesis of load characteristic is presented in this paper. SOM neural network can automatically cluster inputting modes, the trained SOM neural network can discriminate a new inputting mode. Using industry load composition of each load substation to form feature vector,48 substation load characteristics are classified and synthesized by using SOM neural network, the results are satisfactory. Keeping the existing classification result unchanging, the new substation load characteristic can be discriminated by using the trained SOM neural network conveniently, which avoids the expensive cost caused by reclassification. This method decreases the randomness and subjectivity of classification and synthesis for load characteristic, and offers a new way for practical load modeling.